Research On Assistant Diagnosis Based On Multi-modal Breast Cancer Image Fusion

Breast cancer,as one of the common cancers,threatens women’s health at all times.As one of the common imaging techniques,magnetic resonance imaging(MRI)is significant for the early screening of breast cancer.With the rapid development and iterative update of imaging technology,diffusion weighted imaging(DWI)technology,dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)technology and magnetic resonance spectroscopy technology,are also frequently appeared in the early clinical screening of breast cancer,which bringing the gospel to all stages of clinical diagnosis of breast cancer patients.At present,many researchers study multi-modal medical image fusion technology by combining the differences between the lesions in multiple modalities to obtain more lesion information.Relevant medical research shows that breast DWI as auxiliary images of DCE-MRI can effectively improve the accuracy of breast lesion detection and ensure the reliability of breast cancer detection.But for breast cancer,the two modalities of DCEMRI and DWI there are few researches on the fusion of state images.For disposing the above problems,the paper will concentrate on the two modalities of standard breast cancer MRI: DCE-MRI and DWI.By fusing the image data under the two modalities,we can obtain richer lesion information data for benign and malignant classification,and appraise the value of auxiliary diagnosis by using fusion in the diagnosis of breast cancer.The specific research content includes the following three parts,(1)Based on the preprocessing of DCE-MRI and DWI image data.Introduced the operation of image screening,image pairing and image expansion on the original images of the two modalities.After obtaining the region of interest,the geometric and texture features of the DCE-MRI and DWI are extracted respectively.(2)Based on the registration research of DCE-MRI and DWI image data.A registration method based on deep learning is proposed to promote the accuracy of the registration of the two modal images.In the context of traditional registration methods,the VGG-16 network structure is selected as the basic registration network structure,and an iterative VGG-16 network framework is proposed to realize the registration of DCE-MRI and DWI images.The experimental results show that the iterative VGG-16 network structure is more suitable for the registration of DCE-MRI and DWI image data.(3)Based on the fusion research of DCE-MRI and DWI image data.For DCE-MRI and DWI images after registration,this paper uses a combination of featurelevel and decision-level fusion methods to classify breast images.The simple classifier SVM,KNN,and decision tree algorithm are used to classify and compare single-modal images.And put a value on the performance of multi-mode image fusion in assisting diagnosis of breast cancer classification,through analyzing the classification results of comparative experiments.